debayan (@debayan) 's Twitter Profile
debayan

@debayan

Researcher at Leuphana University of Lüneburg, Germany . Knowledge Graphs, Question Answering, Natural Language Processing. Views are personal. 🇮🇳 🇩🇪

ID: 15313431

linkhttp://debayan.github.io calendar_today03-07-2008 23:00:08

22,22K Tweet

1,1K Followers

2,2K Following

François Chollet (@fchollet) 's Twitter Profile Photo

The most important skill for a researcher is not technical ability. It's taste. The ability to identify interesting and tractable problems, and recognize important ideas when they show up. This can't be taught directly. It's cultivated through curiosity and broad reading.

Shubham Saboo (@saboo_shubham_) 's Twitter Profile Photo

China's Alibaba just dropped a Python framework for building multi-agent apps. AgentScope lets you build AI agents visually with MCP tools, memory, rag, and reasoning capabilities. Works with any LLM and supports real-time steering. 100% Opensource.

China's Alibaba just dropped a Python framework for building multi-agent apps.

AgentScope lets you build AI agents visually with MCP tools, memory, rag, and reasoning capabilities.

Works with any LLM and supports real-time steering.

100% Opensource.
debayan (@debayan) 's Twitter Profile Photo

I applied for a Japanese visa recently. The visa officer was like "If you send me an email, I will reply within 24 hours. If I do not, it means our mail server has failed."

Alexia Jolicoeur-Martineau (@jm_alexia) 's Twitter Profile Photo

New paper 📜: Tiny Recursion Model (TRM) is a recursive reasoning approach with a tiny 7M parameters neural network that obtains 45% on ARC-AGI-1 and 8% on ARC-AGI-2, beating most LLMs. Blog: alexiajm.github.io/2025/09/29/tin… Code: github.com/SamsungSAILMon… Paper: arxiv.org/abs/2510.04871

Yuchen Jin (@yuchenj_uw) 's Twitter Profile Photo

Google has won 3 Nobel Prizes in just 2 years: >2024: Demis Hassabis (AlphaFold), Geoff Hinton (AI) >2025: Michel & Jone (Quantum Computing) No other tech company matches Google’s long-term research taste and ambition. Not Apple, Microsoft, or Amazon. The real question: when

Yam Peleg (@yampeleg) 's Twitter Profile Photo

Karpathy just dropped a code that: 1. Trains a tokenizer (rust). 2. Pretrains a GPT model on FineWeb. 3. SFT on chat data (smoltalk) & multiple choice questions & tool use. 4. RL (GRPO) on GSM8K. 5. Evaluates on ARC, MMLU, GSM8K, HumanEval. 6. Run Inference with: KV cache,

debayan (@debayan) 's Twitter Profile Photo

omas-gegen-rechts.org spotted them at a stall in the local market the other day. "Grandmothers against right-wing extremism"